This all seems right. However, if there were no conferences, it would make it hard for geographically isolated researchers. How about just decoupling conferences from publication, like almost all other fields?

That seems both destructive and delusional. Discrimination always exists in all systems. Destroy the system and what remains is nothing, which would be much worse.

Let's not forget that, whatever Trump's motivation, the affected countries refuse to share data about the travelers with US security. Why not put pressure on Iran and the other countries to combat terrorism, rather than use it as a tool?

Many people in developing countries are unable to attend conferences due to cost. This was true and is still mostly true for researchers in India though things have improved substantially in the recent past. The system adapts to that in various ways. Abolishing conferences solely for this reason seems like too much of a step. Especially in TCS I would like to believe that one can have a fine research career without publishing in conferences. It may even be liberating to do that.

I favor open borders but conferences are tremendously expensive... I hardly ever publish in conferences. I don’t feel cut off socially one bit. Whether I’d do better as a scientist with more conference attendance is an open question but I doubt it. I regularly work with some of the best people I can hope to meet.

+Daniel Lemire Is it possible that your case does not model the general case very well? I guess the discussion should probably not be based on particular individuals but should rather raise the question on how processes in science could be designed from a universal perspective, in Kant's sense.
Especially with hiring committees in mind, I do think that there is a significant selection pressure that favors those that can travel a lot, let it be monetary, legally or socially. Is this a good thing?

+Stefan Huber I am not offering myself as a model... but if the computer science model is universal, one should explain to the physicists and biologists that they are doing it wrong.

+Daniel Lemire Oh, I am sorry, I thought you are arguing in a specific direction when you talked about your particular situation. That said, I never argued that the "CS model" (or any other) would be universal for science, which would be obviously wrong.

+Stefan Huber Please, don't copy me. My point is that the idea that you need absolutely to publish in conferences (an expensive and environmentally disastrous approach) for science to move forward is wrong.

+Daniel Lemire I agree, for science it is not. For scientists, in our current system, however, I believe it is kind of, or at least the competitive disadvantage is significant.